CN-122020140-A - Complex equipment semantic information extraction method

CN122020140ACN 122020140 ACN122020140 ACN 122020140ACN-122020140-A

Abstract

The invention relates to the technical field of industrial interconnection, in particular to a semantic information extraction method of complex equipment, which comprises the following steps of firstly, constructing a one-dimensional semantic information extraction model of the complex equipment, adopting a Times Block model, converting data of the complex equipment from a time domain to a frequency domain by utilizing fast Fourier transform, adding wavelet transform before Fourier transform is carried out on the Times Block model to realize the extraction of the semantic information of the one-dimensional data, secondly, constructing a time sequence diagram neural network model for extracting the semantic information among sensors of the complex equipment, and thirdly, constructing a multidimensional semantic information extraction model of the complex equipment on the basis of the first step and the second step. By constructing the time Block model based on wavelet transformation to extract semantic information of one-dimensional data, experimental comparison analysis shows that the classification accuracy of the time Block model integrated with wavelet transformation is higher than that of other methods, and the feature information contained in the data can be effectively captured.

Inventors

XIAO YANJUN
ZENG XUEBIN

Assignees

河北工业大学

Dates

Publication Date: 20260512
Application Date: 20260204

Claims (8)

1. The complex equipment semantic information extraction method is characterized by comprising the following steps of: step one, constructing a one-dimensional semantic information extraction model of complex equipment; The method comprises the steps that a Times Block model is adopted, the Times Block converts data of complex equipment from a time domain to a frequency domain by utilizing fast Fourier transform, and wavelet transform is added before the Fourier transform is carried out on the Times Block model, so that the extraction of semantic information of one-dimensional data is realized; step two, constructing a time sequence diagram neural network model for extracting semantic information among sensors of complex equipment: And thirdly, constructing a multi-dimensional semantic information extraction model of the complex equipment on the basis of the first step and the second step.
2. The complex equipment semantic information extraction method according to claim 1, wherein the wavelet basis function of the wavelet transformation is Symlet wavelet, and the wavelet transformation formula is discretized in the process of computing and solving the wavelet transformation formula.
3. The method for extracting semantic information of complex equipment according to claim 2, wherein in the first step, after obtaining a one-dimensional denoising frequency domain signal, the one-dimensional denoising frequency domain signal is periodically cut and transposed to obtain a two-dimensional tensor, after transformation, the two-dimensional tensor is processed by using a Inception module, and then the learned two-dimensional tensor is converted back into a one-dimensional information space.
4. The method for extracting semantic information of complex equipment according to claim 1, wherein the specific steps of the second step are as follows: Constructing a graph; Constructing a semantic extraction model among complex equipment sensors; Semantic information and relation among parts of complex equipment are obtained by using a sequence chart attention network, node characteristics are converted into vector space, and then characteristic vectors among the nodes are extracted by using a chart attention mechanism.
5. The method for extracting semantic information of complex equipment according to claim 4, wherein the central node obtains the attention coefficient of each node from the feature vectors of the adjacent nodes by using softmax operation, and obtains the attention coefficient, and then performs weighted summation on the attention coefficient to obtain the relevant features of the adjacent nodes, so as to obtain the aggregated features, and the attention mechanism obtains the time relation.
6. The method for extracting semantic information from complex equipment according to claim 4, wherein the second step further comprises calculating by using a multi-head Attention mechanism, wherein the multi-head Attention mechanism comprises a plurality of Attention heads, each head learns different semantic information, the information is integrated and calculated, the calculated results are spliced together through a matrix, a graph Attention layer is formed by the calculation process, the graph Attention layer performs self-Attention calculation according to a node characteristic set of data input, attention is distributed from a node to an adjacent node through Mask Attention masking operation, and an Attention value of the node to the adjacent node is obtained through a feedforward network function.
7. The complex equipment semantic information extraction method according to claim 6, comprising fusing a complex equipment single sensor information semantic information extraction model with a complex equipment sensor relationship semantic information extraction model.
8. The method for extracting semantic information of complex equipment according to claim 6, wherein in the third step, for the data of each node, firstly, the semantic information of each node is acquired by the Times Block module through the Times Block module, after the semantic information of each node is captured, the correlation of the node in space and time is acquired by using an attention mechanism, a graph with correlation in space and time is obtained through calculation of a graph attention network, and the semantic information of the node in the graph contains the correlation with the adjacent node in space and time, and the information of other nodes in the graph is acquired according to the correlation.

Description

Complex equipment semantic information extraction method Technical Field The invention relates to the technical field of industrial interconnection, in particular to a semantic information extraction method for complex equipment. Background Various industries are facing the transformation of traditional communication technologies into intelligent, informationized and digitized communication modes. This transformation is critical to improving transmission efficiency and meeting the future mass data transmission requirements. Currently, most industries mainly adopt traditional communication modes, and these modes are worry when facing challenges of big data age, and major problems include low transmission speed, low automation degree, insufficient security, and difficulty in adapting to the continuous development requirement of future communication. The semantic communication is a communication mode which compresses deep semantic information by capturing the deep semantic information in the original data and transmits the deep semantic information by utilizing information of a semantic layer. Semantic communication differs from traditional communication by the fact that semantic communication is task oriented, understanding and transmitting. Compared with traditional communication, the communication mode remarkably improves the transmission efficiency and reliability of the communication process. Therefore, the research of the semantic information extraction method of the complex equipment has important significance, and is a key for combining the complex equipment with semantic communication to make the complex equipment take an important step on an intelligent road. The most critical step in this process is the acquisition of semantic information in the original data, because whether there is high quality semantic information determines whether the semantic communication system can efficiently transmit data. With the development of artificial intelligence technology in recent years, the acquisition of high-quality semantic information in data becomes possible. Deep learning, which is the most important part in the field of artificial intelligence, has excellent feature extraction capability, and features in data can be learned and applied. The prior art has the following disadvantages: 1. In the industrial field, studies on learning data features by deep learning so as to perform feature extraction are few, and because industrial data has noise and interference in the transmission process, problems such as data loss are easily caused, and the reliability of communication technology is low. 2. In the process of extracting semantic information, the traditional method can only extract the semantic information of single data and can not acquire the associated information existing between different sensor data. Therefore, how to efficiently and accurately extract semantic information in original data of complex equipment to solve the above-mentioned problems is a problem that a person skilled in the art needs to solve at present, Therefore, the application provides a complex equipment semantic information extraction method. Disclosure of Invention In order to make up the defects of the prior art and solve the technical problems in the background art, the invention provides a complex equipment semantic information extraction method. The invention is realized by the following technical scheme: a complex equipment semantic information extraction method comprises the following steps: step one, constructing a one-dimensional semantic information extraction model of complex equipment; The method comprises the steps that a Times Block model is adopted, the Times Block converts data of complex equipment from a time domain to a frequency domain by utilizing fast Fourier transform, and wavelet transform is added before the Fourier transform is carried out on the Times Block model, so that the extraction of semantic information of one-dimensional data is realized; step two, constructing a time sequence diagram neural network model for extracting semantic information among sensors of complex equipment: And thirdly, constructing a multi-dimensional semantic information extraction model of the complex equipment on the basis of the first step and the second step. Preferably, the wavelet basis function of the wavelet transformation is Symlet wavelet, and the wavelet transformation formula is discretized in the process of calculating and solving the wavelet transformation formula. Preferably, in the first step, after a one-dimensional denoising frequency domain signal is obtained, cyclic clipping is performed and transposition is performed to obtain a two-dimensional tensor, after transformation, the two-dimensional tensor is processed by using a Inception module, and then the learned two-dimensional tensor is converted back into a one-dimensional information space. Preferably, the specific steps of the second step are as follow